puda-machines
Pass
Audited by Gen Agent Trust Hub on Apr 17, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
pudaCLI to discover machine commands (e.g.,puda machine commands biologic). This is an expected use of local development tools by the agent to acquire context. - [COMMAND_EXECUTION]: The skill provides instructions to run an edge service for mass balance hardware using
uv run. This command targets local script files and is part of the intended hardware interface. - [EXTERNAL_DOWNLOADS]: Uses
uv syncto manage Python environments and dependencies. This is a standard operation for Python-based automation skills to ensure necessary drivers are available. - [REMOTE_CODE_EXECUTION]: The skill generates Python code for Opentrons robot protocols. This generation is the core functionality and is governed by instructions that require user verification of deck slots and parameters.
- [DATA_EXFILTRATION]: The balance machine driver publishes mass telemetry to a NATS server as part of its monitoring function. This network operation is documented and configurable via environment variables.
- [PROMPT_INJECTION]: The skill processes natural language instructions and CSV data to generate protocols, creating a surface for indirect prompt injection. This is addressed by strict rules requiring user confirmation for machine selection, deck slots, and missing parameters (Ingestion: User requests, CSV files; Boundary: Mandatory confirmation rules; Capabilities: CLI execution, protocol generation; Sanitization: User review).
Audit Metadata